Describe what you need
Use AI Assist to describe the kind of widget you want — an intake form, a status display, a confirmation control — in plain language.
Widgets transform agent messages from text into interactive, structured interfaces. Forms, layouts, and controls — embedded directly in the thread, where the work is happening.
A message is only useful when someone can act on it. Text alone leaves the last mile of every interaction to the person reading it.
The response should be where the work finishes, not where it restarts.
When an agent returns a wall of text, users have to read it, parse it, and then act on it somewhere else. The response is a step, not a destination.
When agents need users to provide information, free-form conversation is inefficient. Structured input — fields, dropdowns, buttons — gets the right data, every time.
When users have to copy information out of a thread and paste it into another system, the workflow breaks. The work should not leave the conversation.
A Widget is a rich, interactive component that an agent embeds directly into a thread message — replacing plain text with structured layouts users can read, fill in, and act on without leaving the conversation.
Widgets are built from a library of interface components. Combine fields, buttons, labels, and layouts to create the experience each agent and workflow requires.
Describe the widget you want in plain language. AI Assist generates the editable code and schema automatically — no building from scratch.
Users see the widget as part of the agent's message and interact with it directly — filling in fields, clicking buttons — without opening another tool or leaving the conversation.
Interacting with a widget can invoke agent actions, including Tool calls. A button click or form submission triggers the agent to act on what the user provided.
Use AI Assist to describe the kind of widget you want — an intake form, a status display, a confirmation control — in plain language.
AI Assist generates the widget and its schema. Review and edit as needed to match your exact requirements.
Attach the widget to an agent. The agent can now embed it into thread messages when the work calls for it.
When the agent sends the widget, users interact with it right where the conversation is happening. Input is captured, actions are triggered, work moves forward.
Structured layouts, forms, and controls replace text-heavy replies — making it clear what a user needs to do and making it easy to do it right there.
When agents need information from users, widgets ensure that input arrives in the right format. No ambiguous free-text answers to interpret.
Plain-language descriptions become working, editable widgets. Builders spend less time on interface code and more time on the work that matters.
Because widgets live in the thread, users provide input, trigger actions, and complete tasks in the same conversation — no context switching required.
Intake, reporting, and approvals — captured as structured input and acted on without leaving the conversation.
A product manager wants their team to submit feature ideas and bug reports in a consistent format. Using AI Assist, they describe a simple intake form. Opal generates the widget. They attach it to their intake agent. When a teammate asks the agent to log an idea, the agent responds with the form in the thread — the teammate fills in the fields, submits, and the agent captures the structured details.
A program manager builds a weekly reporting agent that asks each team lead for a project update. Instead of parsing free-text answers, the agent sends a structured form with defined fields for status, blockers, and next steps. Every response arrives in the same format — ready to roll up into the weekly report.
A finance operations team configures an agent that handles expense and reimbursement requests. When a team member requests a reimbursement, the agent sends a widget with the necessary fields. When the form is submitted, the agent triggers the approval Tool and confirms the submission — all within the thread.
Permissions control who can create, edit, and manage Widgets and attach them to agents.
Widget configuration and agent interactions are tracked in the platform's activity log.
Guardrails apply to agent messages — including those that embed widgets — ensuring all outputs conform to policy.
ZDR is enabled automatically for models and providers that support it, covering the intelligence that powers agent responses.
ISO 27001, ISO 42001, and SOC 2 Type II certifications are underway.